So we've been playing around with comparisons between RNA-seq and RNA FISH, and while we're still a ways away from a more comprehensive study, I thought I'd share a little tidbit. There are two sorts of comparisons one might want to make when quantifying gene expression via RNA-seq. One is comparing expression of a single gene in different conditions; i.e., find the "hits". We have found this to be reasonably good at finding hits, probably with a validation rate of around 50% or more. The magnitude of the hits are, well, sort of close to what the RNA-seq would predict.
The other is to compare the expression of different genes in the same sample. Here, the relevant question is whether genes with similar read counts (FPKM/RPKM) have the same actual number of RNA as measured by RNA FISH. Basically, what's the correlation between RNA-seq and RNA FISH? We've found that the correlation is quite good–but with an important caveat. The correlation is good for high expressing genes. This means that if you have twice as many beta-actin FPKM as GAPDH FPKM, chances are you'll have roughly twice (ish) the actual number of beta-actin as GAPDH mRNA in your cells. However, as you get down to below around 50-100 FPKM, there is much less correlation and way more noise. That means that if you have Transcription Factor A and Transcription Factor B with FPKMs of 25 and 50 respectively, pretty much all bets are off as to whether TF A or TF B will have more actual mRNA numbers. Usually, people seem to care more about the expression of these low genes anyway, so it's important to keep this in mind. Another point is that if you're getting low FPKM values (I'm going to be vague about how low), there's a good chance that the RNA is not actually expressed at all in the cell line you're looking at. We've seen it happen a lot. A LOT. You know who you are... :)
Anyway, hope this information is useful to someone, somewhere. I promise we'll have a lot more on this sometime soon.